evaluation-task · Source-linked discovery
IFEval: Instruction-Following Evaluation
Evaluates how well language models can strictly follow detailed instructions, such as writing responses with specific word counts or including required keywords.
OriginJeffrey Zhou, Tianjian Lu, Swaroop Mishra et al.TopicsGeneral capabilityStatusimported
Can support
Not independently assessed by FronteraEval yet.
Cannot support by itself
No inference beyond the upstream source should be made until the protocol is reviewed.